Regardless if this or other paper is what you end up implementing, I wanted to say kudos for doing the right thing - having the right approach. Building stuff that's just now possible - in AI, from research papers - and productizing it, is how you get ahead of everyone else.